beu（贝乌）

beu -- Session Memory for AI Agents

You have access to beu, a persistent session-memory CLI. Use it to track work across sessions and survive conversation compaction.

When to Use beu

• Work spans multiple sessions or days
• Need to persist decisions, blockers, or context
• Tracking deliverables, tasks, or investigations
• Resuming after conversation compaction

Session Protocol

beu resume              # Check checkpoint, blockers, focus items
beu journal open        # Start interaction ledger
# ... do work ...
beu check               # Verify docs are up to date (if configured)
beu pause "state desc"  # Save checkpoint before stopping

Module Quick Reference

Test Status Tracking

Every task carries a test_status field that tracks the testing lifecycle:

planned -> designed -> implemented -> tested -> darklaunched -> launched

# View test pattern reference (unit, integration, systest, golden)
beu test patterns

# Update test status after writing tests
beu task test-status <id> implemented

# Mark tests as run and passing
beu task test-status <id> tested

# Filter tasks by test status
beu task list --test-status planned

Rules for agents:

• When adding a task for a new feature, test_status starts as planned.
• After deciding what tests to write (consult beu test patterns), set it to designed.
• After writing the tests, set it to implemented.
• After running the tests and confirming they pass, set it to tested.
• Include beu task list --test-status planned in the pre-pause checklist to catch untested work.

Common Patterns

Record a decision

beu state set --category decision "db-choice" "SQLite for simplicity"

Track a blocker

beu state set --category blocker "ci-flaky" "Tests timeout on CI"

Task workflow

beu task add "implement auth" --priority high --tag feature
beu task sprint                    # View sprint board
beu task update 1 --status in-progress
beu task done 1

Test workflow

beu test patterns                  # See available test patterns
beu task test-status 1 designed    # Decided what tests to write
beu task test-status 1 implemented # Tests written
beu task test-status 1 tested      # Tests passing
beu task list --test-status planned # Find tasks still needing tests

Debug investigation

beu debug open "connection timeout"
beu debug symptom conn-timeout "errors after 30s idle"
beu debug log conn-timeout "pool max_idle is 10s"
beu debug cause conn-timeout "idle timeout shorter than keepalive"
beu debug resolve conn-timeout

Artifact tracking

beu artifact add design --type doc
beu artifact status design in-progress
beu artifact changelog design "added auth section"

System Commands

beu status              # Project overview + enabled modules
beu progress            # Cross-module summary
beu events              # Recent event log
beu health              # Database integrity check
beu check               # Compliance gate (required docs)
beu export --all        # Export all data as JSON
beu update-rules        # Re-install latest skill files

Configuration

.beu/config.yml controls which modules are active:

modules:
  journal: true
  artifact: true
  task: true
  state: true
  idea: true
  debug: true

# Optional: override built-in test patterns
test_patterns:
  - key: property
    description: QuickCheck/proptest-based property testing.

Disabled modules return an error when their commands are invoked.

Project Scoping

Use -p <project> to scope commands to a specific project within a shared .beu database. Without the flag, the default project is used.

Global Flags

• --beu-dir <PATH>: Override.beu directory location
• -p, --project <ID>: Project scope
• -v, --verbose: Detailed output
• -q, --quiet: Suppress non-essential output